Vol 456 | 27 November 2008 | doi:10.

1038/nature07389

LETTERS
A fast, robust and tunable synthetic gene oscillator
Jesse Stricker1*, Scott Cookson1*, Matthew R. Bennett1,2*, William H. Mather1, Lev S. Tsimring2 & Jeff Hasty1,2

One defining goal of synthetic biology is the development of decreases promoter activity, and the differential activity of the two
engineering-based approaches that enable the construction of feedback loops can drive oscillatory behaviour1,13.
gene-regulatory networks according to ‘design specifications’ The oscillator cells (denoted JS011) exhibited ubiquitous fluor-
generated from computational modelling1–6. This approach pro- escence oscillations over the entire run time of each experiment (at
vides a systematic framework for exploring how a given regulatory least 4 h). For example, at 0.7% arabinose and 2 mM IPTG, more
network generates a particular phenotypic behaviour. Several fun-
damental gene circuits have been developed using this approach, a + Arabinose b
12

Fluorescence (a.u.)
including toggle switches7 and oscillators8–10, and these have been
applied in new contexts such as triggered biofilm development11 araC yemGFP 8
and cellular population control12. Here we describe an engineered
genetic oscillator in Escherichia coli that is fast, robust and per- − IPTG 4
sistent, with tunable oscillatory periods as fast as 13 min. The
oscillator was designed using a previously modelled network lacI 0
0 60 120 180
architecture comprising linked positive and negative feedback Time (min)
loops1,13. Using a microfluidic platform tailored for single-cell c d
microscopy, we precisely control environmental conditions and
monitor oscillations in individual cells through multiple cycles.
Experiments reveal remarkable robustness and persistence of
oscillations in the designed circuit; almost every cell exhibited
large-amplitude fluorescence oscillations throughout observation
runs. The oscillatory period can be tuned by altering inducer
levels, temperature and the media source. Computational model- 0 60 120 180 0 60 120 180
ling demonstrates that the key design principle for constructing a e f
robust oscillator is a time delay in the negative feedback loop,
which can mechanistically arise from the cascade of cellular pro-
cesses involved in forming a functional transcription factor. The
positive feedback loop increases the robustness of the oscillations
and allows for greater tunability. Examination of our refined
model suggested the existence of a simplified oscillator design
without positive feedback, and we construct an oscillator strain
0 60 120 180 0 60 120 180
confirming this computational prediction.
g h
The synthetic gene oscillator is based on a previously reported
theoretical design1 and was constructed using E. coli components
(Fig. 1a). The hybrid promoter (plac/ara-1; ref. 14) is composed of
the activation operator site from the araBAD promoter placed in
its normal location relative to the transcription start site, and repres-
sion operator sites from the lacZYA promoter placed both upstream
and immediately downstream of the transcription start site. It is
activated by the AraC protein in the presence of arabinose and 0 60 120 180 0 60 120 180
Time (min)
repressed by the LacI protein in the absence of isopropyl b-D-1-thio-
galactopyranoside (IPTG). We placed the araC, lacI and yemGFP Figure 1 | Oscillations in the dual-feedback circuit. a, Network diagram of
(monomeric yeast-enhanced green fluorescent protein) genes under the dual-feedback oscillator. A hybrid promoter plac/ara-1 drives transcription
the control of three identical copies of plac/ara-1 to form three co- of araC and lacI, forming positive and negative feedback loops. b, Single-cell
regulated transcription modules (Supplementary Infor- fluorescence trajectories induced with 0.7% arabinose and 2 mM IPTG.
mation). Hence, activation of the promoters by the addition of ara- Points represent experimental fluorescence values, and solid curves are
smoothed by a Savitsky–Golay filter (for unsmoothed trajectories, see
binose and IPTG to the medium results in transcription of each Supplementary Fig. 3). The trajectory in red corresponds to the density map
component of the circuit, and increased production of AraC in the above the graph. Density maps for trajectories in grey are shown in g. a.u.,
presence of arabinose results in a positive feedback loop that arbitrary units. c–h, Single-cell density map trajectories for various IPTG
increases promoter activity. However, the concurrent increase in conditions (c, 0 mM IPTG; d, 0.25 mM; e, 0.5 mM; f, 1 mM; g, 2 mM;
production of LacI results in a linked negative feedback loop that h, 5 mM).
1
Department of Bioengineering, University of California, San Diego, La Jolla, California 92093, USA. 2Institute for Nonlinear Science, University of California, San Diego, La Jolla,
California 92093, USA.
*These authors contributed equally to this work.
516
©2008 Macmillan Publishers Limited. All rights reserved
NATURE | Vol 456 | 27 November 2008 LETTERS

than 99% of the cells showed oscillations with a period of approxi- period as the cells were imaged on the microfluidic device (Supple-
mately 40 min (Fig. 1b, g, Supplementary Table 1 and Supplementary mentary Fig. 4). This increase was not seen in doubling times, imply-
Movie 1). The highly dynamic nature of the oscillator components is ing that the cells were not experiencing nutritional difficulties
shown by the rapid decay of green fluorescent protein (GFP) signal, or environmental stress that might cause an alteration in oscillator
which drops from peak to trough in less than 10 min (Fig. 1b). The behaviour.
oscillatory phase was heritable between daughter cells, which resulted To explore further the robustness of the oscillator, we investigated
in synchronized oscillations in areas of the microcolony derived from the effect of varying arabinose, temperature and the media source. At a
a common cell. This synchrony was limited to a few periods, pre- fixed value of 2 mM IPTG and at 37 uC, the oscillatory period can be
sumably owing to oscillatory phase diffusion. We used a microfluidic tuned from 13 min to 58 min by varying the arabinose level from 0.1%
device with a laminar boundary switch upstream of the growth to 3.0% (Fig. 2b). Cells grown in the absence of arabinose did not
chamber to investigate the initiation of synchronized oscillations express measurable levels of GFP in single-cell microscopy or flow
(Supplementary Fig. 2c, d). Cells grown in the absence of inducer cytometry experiments, and high levels of arabinose seemed to saturate
initiated oscillations in a synchronous manner on the addition of the system. We observed sustained oscillations at a range of tempera-
inducer (Supplementary Movie 10), which suggested the possibility tures from 25 uC to 37 uC, with a decreasing period as a function of
of using flow cytometry to characterize the oscillator further. Flow temperature (Fig. 2c). The cell doubling time also decreased with
cytometry of samples continuously collected from a culture in loga- temperature, as expected, and the oscillatory period increased mono-
rithmic growth that had been induced with 0.7% arabinose and tonically with cell doubling time (Fig. 2d). The oscillator also func-
2 mM IPTG showed oscillations in mean cell fluorescence tioned in minimal A medium with 2 g l21 glucose (Fig. 2c, d).
(Supplementary Fig. 8). Induction of oscillation was very quick (less Although the cell doubling time in minimal medium was significantly
than 5 min) and initially well-synchronized. The amplitude of these longer than in LB-Miller formulation lysogeny broth (LB) (80–90 min
bulk oscillations decayed as the experiment progressed, as expected versus 22–24 min at 37 uC), the period in the minimal medium was
from the desynchronization of individual cells in the colony very similar to that in LB (Fig. 2c, d). This result, together with the
(Supplementary Information). However, the period obtained from strong dependence of the period on IPTG and arabinose concentration
the flow cytometry method (green data points in all figures) com- (at constant cellular doubling times), demonstrates that the synthetic
pared favourably to that obtained from single cells using microscopy oscillator is not strongly coupled to the cell cycle. The similar depend-
(red data points in all figures). ence of the period and the doubling time on the temperature seems to
The oscillator was extremely robust over an extensive range of be due to the thermodynamic change of the rate constants affecting all
inducer conditions and temperatures. At 0.7% arabinose and 37 uC, cellular processes.
almost every observed cell oscillated (Supplementary Table 1) at all The oscillator was constructed according to design principles deter-
IPTG concentrations examined (Fig. 1b–h and Supplementary mined from previous theoretical work1. However, we found that this
Movies 1–8). Varying the IPTG concentration allowed for the tuning original model failed to describe two important aspects of the experi-
of the oscillator period (Fig. 2a), particularly at low IPTG concentra- ments. First, the model could not describe the observed functional
tions. The period decreased at high IPTG concentrations, and sub- dependence of the period on inducer levels. Second, and perhaps most
sequent characterization of the promoter revealed that this non- importantly, because careful parameter tuning was necessary for
monotonic behaviour is probably caused by IPTG interference with oscillations in the original model, it was not able to describe the robust
AraC activation15 (Supplementary Information). The cell doubling behaviour demonstrated in the experiments. This suggests that only a
time on the microfluidic device remained largely steady between small region of inducer space should support oscillations, in contrast
experiments, ranging from 22.3 min to 27.6 min at 37 uC and showing to the robust behaviour demonstrated in the experiments. These
little correlation to IPTG concentration (R2 5 0.132). Individual cell shortcomings forced a re-evaluation of the derivation of the oscilla-
fluorescence trajectories showed a gradual increase in oscillatory tor equations, and led to a new computational model that more
accurately described the experimental observations. The new model
60 60 incorporates the same coupled positive and negative feedback archi-
a 37 °C (G = 22–24 min) b tecture, but includes details that were omitted from the previous
0.7% arabinose
model. In particular, we found that directly modelling processes such
as protein–DNA binding, multimerization, translation, DNA loop-
Oscillatory period (min)

Oscillatory period (min)

37 °C (G = 22–24 min)
ing, enzymatic degradation and protein folding greatly increased the
2 mM IPTG accuracy of the model. The result is a computational model that is very
0 0 robust to parameter variations and correctly describes the dynamics
0 10 20 30 0 1 2 3
IPTG (mM) Arabinose (%) of the oscillator for a large range of IPTG and arabinose concentra-
120 120 tions (see Box 1 and Supplementary Information).
c d
In examining our refined model, we discovered another region in
parameter space that would support oscillatory behaviour. Our
model predicted that a constantly activated system with repression
0.7% arabinose 0.7% arabinose controlled by a negative feedback loop could produce oscillations in
2 mM IPTG 2 mM IPTG the absence of positive feedback (Supplementary Fig. 19). It has been
0 0
20 30 40 0 60 120 proposed that negative feedback gene networks can oscillate as long
Temperature (°C) Cell doubling period (min)
as there is delay in the feedback16,17, and, although there is no explicit
Figure 2 | Robust oscillations. a–c, Oscillatory periods on transects with delay in our model, the intermediate steps of translation, protein
0.7% arabinose and varying IPTG (a), 2 mM IPTG and varying arabinose folding and multimerization of LacI provide an effective form of
(b), or 0.7% arabinose, 2 mM IPTG, and varying temperature (c). Mean delay18 that is sufficient to support oscillations. We constructed this
periods from single-cell microscopy (red diamonds, mean 6 s.d.) or flow
system (denoted JS013) in E. coli using a hybrid promoter, pLlacO-1
cytometry (green circles) are shown. Black curves are trend lines in a and
b, or represent the theoretical prediction based on reference values at 30 uC
(ref. 14), that is activated in the absence of LacI (or presence of IPTG)
in c (see Supplementary Information). Samples grown in minimal medium to drive both lacI and yemGFP expression (Fig. 3a). We observed
rather than LB are indicated by crosses. G represents the cell doubling oscillations in these cells when examined by single-cell microscopy
period. d, Oscillatory period and cell division time increase monotonically as under inducing conditions (Fig. 3b, Supplementary Fig. 5 and
the growth temperature decreases. Symbols are as described above, and the Supplementary Movie 11). These oscillations were not as distinct
black line is a linear regression of samples grown in LB. or regular as in the dual-feedback oscillator, and they did not always
517
©2008 Macmillan Publishers Limited. All rights reserved
LETTERS NATURE | Vol 456 | 27 November 2008

Box 1 | Dynamic modelling of the dual-feedback oscillator circuit

We used standard techniques to construct both stochastic and a Unfolded Folded LacI LacI
deterministic computational models3,25–28 based on the same protein monomer LacI dimer tetramer
underlying biochemical reactions illustrated in Fig. 4a (see mRNA
Supplementary Information for full details of modelling). Although the
interaction between transcription factors and the DNA is generally lacI
quite complicated to model in detail29, we used experimental induction
curves to calibrate the induction levels in the reactions describing the araC mRNA
network (Supplementary Fig. 10). Over many oscillatory cycles, the Unfolded Folded AraC dimer
deterministic simulations were then shown to give accurately the protein AraC
temporal evolution of the mode of the distributions generated by the monomer
exact stochastic simulations24. Representative time series for the
protein concentrations obtained from the stochastic and deterministic b c

concentration
Normalized
models are depicted in Fig. 4b, c. The models are very robust in that 12 1.0

Molecules
(×1,000)
oscillatory behaviour exists for a large range of parameter values and 0.5
6
network details (Supplementary Information). Importantly, we found
excellent quantitative agreement with the experimentally obtained 0 0
0 80 160 0 30 60
period as a function of inducer levels (Fig. 4d, e). Time (min) Time (min)
The amplitude and period of the oscillations as a function of inducer d 50 e
levels can be conceptually explained using Fig. 4c. A burst begins with 60
the basal transcription of messenger RNA from both promoters,

Period (min)

Period (min)
encoding both the activator and the repressor. After a short delay
(caused by, for example, translation, protein folding and 30
multimerization), the amount of functional activator rises quicker than
the amount of functional repressor, as shown in Fig. 4b. This occurs for
two reasons. First, the activator gene is on a higher copy number 10
plasmid than the repressor gene, meaning that more activator 10
0 10 20 30 0 1 2 3
transcripts are produced than repressor transcripts. Second, assuming IPTG (mM) Arabinose (%)
that transcription and translation of the monomeric forms of both
proteins occur at similar rates, the activator will be more abundant Figure 4 | Modelling the genetic oscillator. a, Intermediate processes are
because the functional tetrameric form of LacI requires twice as many explicitly modelled in the refined oscillator model. b, c, Simulation results
monomers as does the functional dimeric form of AraC. As AraC levels from Gillespie simulations (b) or deterministic modelling (c) at 0.7%
rise, an activation burst in production of mRNA occurs due to the arabinose and 2 mM IPTG. AraC dimers (green), LacI tetramers (red) and
positive feedback loop. After LacI has been converted to a sufficient lacI mRNA (black) are shown. d, e, Comparison of modelling and
number of tetramers, the production of mRNA is turned off and the experiment for oscillation period at 0.7% arabinose (d) or 2 mM IPTG
proteins decay enzymatically. Once all proteins have decayed, the (e). Values from deterministic modelling (blue curve), stochastic
promoters are freed of all bound regulators and the cycle begins anew. simulations (grey symbols, Supplementary Fig. 18), and microscopy (red
The length of the period is primarily determined by the time required diamonds) or flow cytometry (green circles) are shown. Lower and upper
for the proteins to decay. Therefore, the period is dependent on the rate error bars in d represent the 16th and 84th percentiles, respectively, of the
of enzymatic decay and the magnitude of the activation burst. stochastic data, corresponding to 61 s.d. for a normal distribution.
Furthermore, because the burst size depends on the induction
characteristics of the promoter, it follows that the period is roughly
proportional to the induction level of the promoter. design of engineered gene circuits. We found that a full model of the
system that takes into account intermediate steps such as multimeri-
zation, translation, protein folding and DNA looping is essential. The
return to a dim state, consistent with the predictions of the computa- reason for this lies not only in the timescales of the system but also in
tional model. Furthermore, the period was largely unaffected by the sequential timing of events. Because the intermediate steps in the
IPTG concentration (varying less than 5% over three experimental production of functional protein take time, their introduction into the
runs from 0.6 mM to 20 mM IPTG), suggesting that the addition of model creates an important form of delay18–20. We found that this
the positive feedback loop serves the dual role of regularizing oscilla- effective delay greatly increases the robustness of our model. For
tions and allowing tunability of the period (see Supplementary instance, oscillatory activity in the model is only somewhat sensitive
Information). to the values chosen for system parameters (Supplementary
In the context of synthetic biology, our findings indicate that cau- Information), implying that nearly all cells should oscillate
tion must be exercised when making simplifying assumptions in the (Supplementary Table 1) despite minor stochastic variations in their
intrinsic parameters. This determination of gene circuit design criteria
in the present context of a fast, robust and tunable oscillator sets the
a b stage for the design of applications such as expression schemes that are
lacI capable of circumventing cellular adaptability, centralized clocks that
pLlacO-1
coordinate intracellular behaviour, and reverse-engineering plat-
− IPTG forms21 that measure the global response of the genome to an oscil-
latory perturbation.

yemGFP METHODS SUMMARY

0 60 120 180 The dual-feedback oscillator circuit was constructed by placing araC, lacI and
pLlacO-1 Time (min)
yemGFP under the control of the hybrid plac/ara-1 promoter14 in three separate
Figure 3 | An oscillator with no positive feedback loop. a, Network diagram transcriptional cassettes. An ssrA degradation tag22 was added to each gene to
of the negative feedback oscillator. This oscillator is similar to the dual- decrease protein lifetime and to increase temporal resolution. These transcrip-
feedback oscillator except that the hybrid promoter pLlacO-1 (ref. 14) gives tional cassettes were placed on two modular plasmids14 and co-transformed into
expression of lacI and yemGFP in the absence of LacI or in the presence of an DaraC DlacI E. coli strain. The negative feedback oscillator circuit was con-
IPTG without requiring an activator. b, Single-cell density map trajectories structed by placing ssrA-tagged lacI and yemGFP under the control of the pLlacO-1
for cells containing this oscillator (see Supplementary Movie 11 and promoter14 in two separate transcriptional cassettes, which were incorporated
Supplementary Fig. 5). onto two modular plasmids and co-transformed into a DlacI strain. Cells were
518
©2008 Macmillan Publishers Limited. All rights reserved
NATURE | Vol 456 | 27 November 2008 LETTERS

grown either in LB medium or in minimal A medium with 2 g l21 glucose. 15. Lee, S. K. et al. Directed evolution of AraC for improved compatibility of arabinose-
Oscillations were induced using arabinose (0.1–3%) and IPTG (0–30 mM). and lactose-inducible promoters. Appl. Environ. Microbiol. 73, 5711–5715 (2007).
Single-cell microscopic data were collected by loading induced cells into 16. Bliss, R. D., Painter, P. R. & Marr, A. G. Role of feedback inhibition in stabilizing the
polydimethylsiloxane-based microfluidic platforms that constrained the cells classical operon. J. Theor. Biol. 97, 177–193 (1982).
17. Bratsun, D., Volfson, D., Tsimring, L. S. & Hasty, J. Delay-induced stochastic
to a monolayer while supplying them with nutrients23, and then providing a
oscillations in gene regulation. Proc. Natl Acad. Sci. USA 102, 14593–14598
constant source of medium and inducers and imaging GFP fluorescence every (2005).
2–3 min for at least 4–6 h. These data were further analysed using ImageJ and 18. Rateitschak, K. & Wolkenhauer, O. Intracellular delay limits cyclic changes in gene
custom-written Matlab scripts to extract single-cell fluorescence trajectories. expression. Math. Biosci. 205, 163–179 (2007).
Flow cytometry was performed either by taking samples from a continuously 19. Mackey, M. & Glass, L. Oscillation and chaos in physiological control systems.
grown and serially diluted culture or by growing multiple cultures in parallel for Science 197, 287–289 (1977).
varying durations. In either case, samples were read directly from their growth 20. Jaeger, J. & Reinitz, J. On the dynamic nature of positional information. Bioessays
medium and low-scatter noise was removed by thresholding. Flow cytometry 28, 1102–1111 (2006).
oscillatory periods were defined as the time elapsed between the first and second 21. Faith, J. et al. Large-scale mapping and validation of Escherichia coli transcriptional
fluorescence peaks. Details of the models discussed are presented in regulation from a compendium of expression profiles. PLoS Biol. 5, e8 (2007).
Supplementary Information. Stochastic simulations were performed using 22. Andersen, J. B. et al. New unstable variants of green fluorescent protein for studies
of transient gene expression in bacteria. Appl. Environ. Microbiol. 64, 2240–2246
Gillespie’s algorithm24, and deterministic simulations were performed using
(1998).
custom Matlab scripts. 23. Cookson, S., Ostroff, N., Pang, W. L., Volfson, D. & Hasty, J. Monitoring dynamics
Full Methods and any associated references are available in the online version of of single-cell gene expression over multiple cell cycles. Mol. Syst. Biol. 1,
the paper at www.nature.com/nature. 2005.0024 (2005).
24. Gillespie, D. T. Exact stochastic simulation of coupled chemical-reactions. J. Phys.
Received 9 July; accepted 5 September 2008. Chem. 81, 2340–2361 (1977).
Published online 29 October 2008. 25. Hasty, J. et al. Computational studies of gene regulatory networks: in numero
molecular biology. Nature Rev. Genet. 2, 268–279 (2001).
1. Hasty, J., Dolnik, M., Rottschäfer, V. & Collins, J. J. Synthetic gene network for 26. Ozbudak, E., Thattai, M., Lim, H., Shraiman, B. & van Oudenaarden, A.
entraining and amplifying cellular oscillations. Phys. Rev. Lett. 88, 148101 (2002). Multistability in the lactose utilization network of Escherichia coli. Nature 427,
2. Hasty, J., McMillen, D. & Collins, J. J. Engineered gene circuits. Nature 420, 737–740 (2004).
224–230 (2002). 27. Wang, X., Hao, N., Dohlman, H. & Elston, T. Bistability, stochasticity, and
3. Tyson, J., Chen, K. & Novak, B. Sniffers, buzzers, toggles and blinkers: dynamics of oscillations in the mitogen-activated protein kinase cascade. Biophys. J. 90,
regulatory and signaling pathways in the cell. Curr. Opin. Cell Biol. 15, 221–231 (2003). 1961–1978 (2006).
4. Sprinzak, D. & Elowitz, M. B. Reconstruction of genetic circuits. Nature 438, 28. Bennett, M. et al. Metabolic gene regulation in a dynamically changing
443–448 (2005). environment. Nature 454, 1119–1122 (2008).
5. Endy, D. Foundations for engineering biology. Nature 438, 449–453 (2005). 29. Gerland, U., Moroz, J. & Hwa, T. Physical constraints and functional
6. Andrianantoandro, E., Basu, S., Karig, D. K. & Weiss, R. Synthetic biology: new characteristics of transcription factor-DNA interaction. Proc. Natl Acad. Sci. USA
engineering rules for an emerging discipline. Mol. Syst. Biol. 2, 2006.0028 (2006). 99, 12015–12020 (2002).
7. Gardner, T. S., Cantor, C. R. & Collins, J. J. Construction of a genetic toggle switch
in Escherichia coli. Nature 403, 339–342 (2000). Supplementary Information is linked to the online version of the paper at
8. Elowitz, M. B. & Leibler, S. A synthetic oscillatory network of transcriptional www.nature.com/nature.
regulators. Nature 403, 335–338 (2000).
Acknowledgements We thank H. Bujard, C. Yang, and Z. Zhang for gifts of
9. Atkinson, M. R., Savageau, M. A., Myers, J. T. & Ninfa, A. J. Development of
reagents, and D. Volfson and M. Simpson for discussions. This work was supported
genetic circuitry exhibiting toggle switch or oscillatory behavior in Escherichia coli.
Cell 113, 597–607 (2003).
by grants from the National Institutes of Health (GM69811-01) and the US
10. Fung, E. et al. A synthetic gene-metabolic oscillator. Nature 435, 118–122 (2005).
Department of Defense.
11. Kobayashi, H. et al. Programmable cells: interfacing natural and engineered gene Author Contributions J.S. and J.H. designed the oscillator circuits, and J.S.
networks. Proc. Natl Acad. Sci. USA 101, 8414–8419 (2004). constructed the circuits. S.C. performed the microscopy experiments, and J.S. and
12. You, L., Cox, R. S., Weiss, R. & Arnold, F. H. Programmed population control by S.C. performed the flow cytometry experiments. S.C., L.S.T. and J.H. performed the
cell-cell communication and regulated killing. Nature 428, 868–871 (2004). single-cell data analysis. M.R.B., W.H.M. and L.S.T. performed the computational
13. Barkai, N. & Leiber, S. Circadian clocks limited by noise. Nature 403, 267–268 modelling. All authors wrote the manuscript.
(2000).
14. Lutz, R. & Bujard, H. Independent and tight regulation of transcriptional units in Author Information Reprints and permissions information is available at
Escherichia coli via the LacR/O, the TetR/O and AraC/I1-I2 regulatory elements. www.nature.com/reprints. Correspondence and requests for materials should be
Nucleic Acids Res. 25, 1203–1210 (1997). addressed to J.H. ([email protected]).

519
©2008 Macmillan Publishers Limited. All rights reserved
doi:10.1038/nature07389

METHODS by flowing heated water through deep thermal channels fabricated into the device.
Oscillator plasmid and strain construction. The oscillator components araC and Cells that had been passed from an overnight culture into inducing media
lacI and a fluorescent reporter protein (yemGFP) were tagged with carboxy- approximately 3–4 h earlier were loaded into the device from the cell port by
terminal TSAANDENYALAA ssrA tags22. yemGFP contains F64L/S65T/A206K directing high flow both from the cell port and from the media port to the waste
mutations. These tagged genes were then cloned into pZ modular plasmids under port. On trapping a single cell in each channel, flow past the cell chamber was
the transcriptional control of the plac/ara-1 hybrid promoter14 to form three co- reversed and slowed to 1–2 mm s21 such that fresh nutrients were delivered from
regulated transcriptional modules with identical promoters, ribosome-binding the media port by means of a combination of diffusion and advection without
sequences and downstream terminators. The plac/ara-1 promoter is activated by physically disturbing the cells.
AraC in the presence of arabinose and repressed by LacI in the absence of IPTG. Cells grew logarithmically to fill the channels over an experimental duration of
The activator araC module and the reporter yemGFP module were placed on a ,4–6 h, while images were acquired every 2–3 min at 3100 magnification in the
ColE1 plasmid, and the repressor module was placed on a p15A plasmid. All PCR- transmitted and fluorescent channels. Focus was maintained during image
amplified sections and sequence junctions were confirmed by sequencing. (See acquisition either by manual adjustment or by contrast-based autofocus algo-
Supplementary Fig. 1.) An DaraC DlacI strain was constructed by P1vir phage rithms. After each imaging session, fluorescence trajectories of individual cells
transduction between DaraC and DlacI strains. The two plasmids described above were extracted using the WCIF ImageJ cell analysis package. For each fluor-
were co-transformed into this strain to construct the dual-feedback oscillator strain. escence frame, mean values of integrated fluorescence were calculated within
To construct the negative feedback oscillator strain, the hybrid promoter constant circular areas inscribed within the boundaries of all tracked cells. Long-
pLlacO-1 (ref. 14) was used to regulate expression of lacI and yemGFP. This term fluorescence trajectories were subsequently constructed by manually track-
promoter is repressed by LacI in the absence of IPTG. Both genes were tagged ing each cell throughout the experiment.
with ssrA tags as described above. Two transcriptional modules containing Flow cytometry. Oscillator cells were initially characterized by flow cytometry of
pLlacO-1 and lacI or yemGFP were constructed as above. The repressor module batch cultures to identify inducer conditions that supported oscillations.
was placed on a p15A plasmid and the reporter module was placed on a ColE1 Subsequently, time-course flow cytometry was performed on growing cultures
plasmid. These were then co-transformed into a DlacI strain. immediately after induction to follow oscillation dynamics. This time-course flow
Microscopy. We examined cells with single-cell time-lapse fluorescence micro- cytometry followed one of two similar protocols. In the continuous protocol, a
scopy using microfluidic devices designed to support growth of a monolayer of single culture was serially diluted to maintain logarithmic growth. The culture was
E. coli cells under constant nutrient flow (Supplementary Fig. 2). The design of induced at the initial time point, and samples were removed for flow cytometry
the microfluidic device used in all microscopy experiments was adapted from the over the course of the experiment. In the aggregate protocol, an uninduced culture
Tesla microchemostat design23 for use with Saccharomyces cerevisiae. in logarithmic growth was aliquoted onto different inducer concentrations, and
Modifications made to support imaging monolayers of E. coli included lowering these subcultures were allowed to grow for varying lengths of time before flow
the cell chamber height to match the cylindrical diameter of K-12 MG 1655 cells, cytometry. Flow cytometry was performed directly from growing cultures, and
lowering the delivery channel height to maintain equivalent flow splitting noncellular low-scatter noise was removed by thresholding. Oscillations were
between the cell chamber and the bypass channel, and dividing the cell trapping tracked by measuring the mean cellular fluorescence at each time point. The
region into three channels for simultaneous observation of isolated colonies amplitude of the initial oscillation was usually higher than that of subsequent
(Supplementary Fig. 2a, b). For on-chip induction experiments, we used a vari- oscillations, presumably owing to desynchronization of the oscillations
ant of this device that incorporated a laminar boundary media switch into the (Supplementary Fig. 8). The oscillation period was defined as the time elapsed
design30 and supported cell growth for several generations in non-inducing between the first and second oscillation peaks. All flow cytometry analysis was
media before induction and imaging (Supplementary Fig. 2c, d). carried out on a Becton-Dickinson FACScan.
In each experiment, a microfluidic device was mounted to the stage and wetted
using a solution of 0.1% Tween 20 surfactant in the appropriate growth medium. 30. Groisman, A. et al. A microfluidic chemostat for experiments with bacterial and
For optimal E. coli growth, the chip temperature was typically maintained at 37 uC yeast cells. Nature Methods 2, 685–689 (2005).

©2008 Macmillan Publishers Limited. All rights reserved